2019 Group Long Term Disability Experience Study Preliminary Report

Question 1

Purpose of the Study

Answer 1

1. Main goal – make recent LTD claim termination data available to companies
   - Also wanted to reduce or eliminate most time-consuming steps in prior study and reduce work required by contributors, SOA and the vendor
2. Secondary objective – document key processes and decisions of study – help future studies be faster and
   more accurate
3. Data request in October 2018 from Group Long-Term Disability (GLTD) Experience Committee
   - 22contributorswith2millionclaims o LargestLTDstudytodate
   - Data from 2009-2017
4. Provides significantly greater volume of data and includes new data about Social Security awards
5. Consolidated Database (CDB)
   - Primary deliverable of study
   - 46segmentationvariables
   - Not given in prior studies – main deliverable was set of pivot tables
   - One record for each month of exposure for each claim (max of 108 month duration per claim)
   ▪ Prior study weighted claims to dampen impact of large companies
   ▪ Makes data more efficient and flexible
   - Contains 47M records
6. This report intended to facilitate users’ ability to access and analyze data from the study
   - Limited amount of analysis on results so far – mostly to confirm accuracy and initial trend analysis

Question 2

Overview

Answer 2

1. Background
   - Options for Future Studies considered by GLTD Committee – Feb – July 2018
   ▪ Enhanced benefit offset study
   ▪ LTD incidence study
   ▪ Update to 2005-2012 LTD claim termination study (decided upon this option)
2. Scope
   - Third option was chosen, and decided to also request limited information on Social Security
   awards
   - Key Objectives for Implementation of Study
   ▪ Complete full cycle (data request to delivery of results) within one year
   ▪ Data request identical to prior 2016 study (plus two SS data fields)
   ▪ Contributor process as close as possible to prior study process
   ▪ Smoother submission process
   ▪ More robust set of self-audit tests
   ▪ Deliverable would be Consolidated Database (CDB)
   ▪ Study to serve as proof of concept for future LTD claim termination studies regarding timeline and budget

• Data request sent in October 2018
• Delivery target date – July 31, 2019– wasn’t met
  ▪ Largely due to inadequate documentation of prior processes, so this should be better in the future

Question 3

Study Definitions

Answer 3

1. Data Submission
   - Companies ask to submit any claims from 2009-2017 that had at least one benefit payment o External vendor (MIB) used to collect and sort data

• Assign Claim Termination to Categories
  ▪ 1. Recovery
  ▪ 2. Death
  ▪ 3. Max Out (max benefit period reached)
  ▪ 4. Limit (expired due to internal benefit limit)
  ▪ 5. Settlement
• Recovery – refers to any termination other than #2-5
  ▪ Not strictly due to recovery from disability (e.g. may be due to change in definition from own-occupation to any-occupation)
• Didn’t examine Max Out, Limit and Settlements in detail
• Certain claims excluded – data from Administrative Services Only claims (ASO), claims from reserve buy-outs, international claims, claims with extended elimination periods (over 15 months) or zero-day elimination periods

2. Segmentation Variables
   - 2016 LTD study – 24 segmentation variables (15 single value variables and 9 range-grouped variables)
   ▪ 2019 study – has same 24 variables plus SS variables

• In addition, has single value data fields, so 46 segmentations variables in total o Sample list of variables [full list can be found in the actual study note]
  ▪ Company Size, Elimination Period, Annual Duration, Age at Disability, Gender, Benefit Amount, Region, Replacement Ratio, SS Award Date, etc

3. Actual-to-Expected Calculations
   - Based on Exposure Basis, Duration, Effective Elimination Period (EEP), and Expected Basis

• Effective Elimination Period (EEP)
  ▪ Benefit commencement date minus date of disability
  ▪ May differ from regular EP due to temporary return to work during EP
• Exposure Basis
  ▪ Ends with earlier of claim termination or end of study period
  ▪ All claims given full month of exposure for each month except:
• Claims on benefit at start of study period, claims on benefit at end of study period. Fractional exposure = Days Exposed / 30 (max 100%)
• ClaimDuration
  ▪ Calendar months from the date of disability
  ▪ Based on Benefit Commencement Date
  ▪ Number of months in EP added on to produce the normal claim duration
• Expected Termination Calculations
  ▪ Based on GLTD2008 tables (same as 2016 study)
  ▪ Termination expectations based on 7 sub-tables from 8 core variables
  ▪ Expected terminations for Max Out, Limit and Settlements not developed

Question 4

Study Process

Answer 4

1. Data Submission
   - Data request sent to all significant writers of LTD in the US
   - Self-audit guide provided to identify specific data integrity checks before submitting data
2. Logic and Syntax Testing
   - MIB (data collection company) created more extensive list of data validation reports
   ▪ Data mostly clean, though still a material number of potential issues to be addressed

• Group devised work-arounds for data issues – objective to produce timely and efficient results, so didn’t want to go back and forth to contributors of data
  ▪ Work-arounds would correct, modify, ignore or delete data fields
  E.g. Carrier didn’t follow proper instructions but clear what they meant to do; Benefit commencement date blank or obviously incorrect – set to be Date of Disability plus EP; Claims terminating within 45 days of max benefit duration date were reclassified as Max Outs
• 8 Core Variables to calculate Expected Terminations
  ▪ Gender, Age at Disability, Duration, EEP, Diagnosis, Gross Indexed Monthly Benefit Amount, Limited Own-Occ Period, and Own-Occ to Any-Occ Transition Month
• Only about 900 claims of the 2M records were deleted
• Carriers not informed of data changes directly but did receive revised submitted data and CDB
• Documented tests and work-arounds to improve efficiency in the future
  ▪ Most could be address by:
• Clearer instructions in Data Request
• Expanding the self-audit
• Clarifying the self-audit instructions

3. Data Validation
   - Types of Data Review in 2019 LTD Study
   ▪ 1. Large company performed independent calculation of exposures and A:E results and matched MIB’s calculations
   ▪ 2. Committee reviewed reports comparing A:E results by contributor across range of variables
   * Variables: Exposure Duration, Claim Duration, Gross Benefit Amount, Taxability Status, STD Integration, Gender, Attained Age, Calendar Year, Calendar Year of Disability, Age at Disability, Case Size, Gender, COLA, Diagnosis Termination Rate, Percent of Claims with SS by Annual Duration
   ▪ 3. Work group reviewed data for reasonableness – analyzed A:E trends over key variables
   ▪ 4. Compared 4 yr overlap period to prior study
4. Data Base Construction
   - 2019 – one record for each month of exposure for each claim (up to 108 months per claim)
   ▪ Significant change from 2016 study where records were weighted to dampen impact of large companies
   ▪ Due to industry consolidation and other industry trends, Committee believed dampening is no longer necessary

• • Subject to rigorous review to make sure Personal Identifiable Information (PII) protected

Question 5

Study Deliverables

Aggregate CDB and several pivot tables used to perform bais analyses

Answer 5

1. CDB Format
   - Includes all variables from 2016 study, plus new variables for SS awards
   - Contains over 47M records
2. Pivot Table Formats
   - Developed several pivot tables for companies to use for basic analysis without having to use the CDB
   ▪ 5 different pivot tables with different segment variables to consider
   * All include Total Monthly Exposure, Avg Gross Benefit, Actual and Expected Recoveries and Deaths, Actual Termination Rates for Settlements, Limits and Max Outs

Question 6

Preliminary Observations on Experience

Answer 6

1. Preliminary Observations on Trend
   a. A:E Ratios by Year
   ▪ A:E recovery ratios – increased every year from 2009 (105.9%) to 2017 (141.0%); Avg
   125%
   ▪ A:E death ratios - decreased almost every year from 2009 (89.4%) to 2017 (82.5%); Avg 86%
   * Shows trends in mortality improvement

b. A:E Recovery Ratios by Integration Type and Claim Duration
▪ Significantly higher for integrated claims (with STD) in early durations
* Could be due to more marginal claims that terminate quickly in integrated segment

c. A:E Recovery Ratios by Diagnosis and Calendar Year
▪ All diagnosis categories (except Maternity) had significant improvements in recovery experience
* Maternity relatively stable (and close to 100%)

d. A:E Death Ratios by Diagnosis and Calendar Year
▪ Significant improvement in Cancer and Diabetes category
▪ Overall A:E death ratio = 86.2%, but many greater than 100% for certain diagnosis categories

e. A:E Death Ratios by Case Size
▪ A:E recovery ratios generally increase with increasing case size
* May be due to workplace accommodations at larger employers to facilitate return to work
* May also reflect severity of claims at larger groups (more marginal) vs smaller groups (more severe)

f. Higher portion integrated with STD – those tend to be more marginal and terminate sooner
▪ A:E death ratios slightly decrease with increasing case size

g. A:E Recovery Ratios by Duration and Calendar Year
▪ Generally increased in every duration group

h. A:E Death Ratios by Duration and Calendar Year
▪ Most significant improvements in mortality in early durations

i. A:E Recovery Ratios and Death Ratios by Company Size and Calendar Year
▪ Generally recovery and death ratios higher for larger sized companies

2. Analysis of 2016 Study and 2019 Study Overlap Period
   a. Underlying exposure differences in 2016 and 2019 studies – slightly different mix of companies, differences in approach to dampen experience

b. Overall Trend by Calendar Year
▪ Trend in A:E Recoveries and Deaths – similar with steady increases in Recovery and steady
decreases in Death rates
* A:E Recovery average – 2016 (113.9%) vs 2019 (125.2%)
* A:E Death average – 2016 (90.6%) vs 2019 (86.2%)
▪ Recovery and Death rates lower in 2019 study on overlapping years
▪ Fitted regression lines to this data
* Recovery rate slope in 2019 greater than 2016
* Death rate slopes were similar

c. Exposure Distribution Differences
▪ By Company Size – exposure in both studies weighted mostly towards Large group, but more companies in Large group in 2019 study
▪ By Diagnosis Grouping – similar across most categories (less Unknown in 2019)
▪ By STD Integration Status – Data more complete in 2019 study; improves ability to analyze impact of STD on LTD recovery and death rates

d. Comparison of Recovery and Death Rates by Duration from Disability Date
▪ 2019 study does not show decline in older claim durations of these rates (as 2016 study did)
▪ Impact of STD integration on recovery rates reduced in 2019 study

3. Analysis of Social Security Status
   - Increasing award percentages by duration
   - Recoveries much lower for claimants with SS awards
   - Deaths much higher for claimants with SS awards

Question 7

Observations on the Use of Study Data

Answer 7

1.Objective was to deliver accurate and usable data in timely fashion
- Didn’t perform depth of analysis needed to conclusively validate data or interpret results

2. A/E Recovery Ratios Through the Own-Occupation/Any-Occupation Transition Period
   - Significant variation in impact of definition of disability in last 3 LTD termination studies
   ▪ 2008 study shows large increase in recoveries in month of change of definition and then declining in the 8 months after the change (but still above normal)
   ▪ Patterns vary due to carrier differences in claim and data processing
   ▪ 2019 study shows elevated recoveries, limit terminations and Max Outs
3. Termination Category for Claims with Internal Limits
   - In addition to claims closing due to internal limit, also increase in Max Outs and recoveries in the limit month and next several months
   ▪ Decided to analyze total terminations (excluding deaths) within 3 months of limit date
   * 2008 study, 2004-2012 study and 2009-2017 study all showed this
   * Mental and Nervous claims recommended to use total terminations (described above)
4. Taxability of Benefits
   - Compare A:E by Full Taxable, Partially Taxable, Not Taxable and Unknown status
   ▪ 2016 study – unusual results – highest A:E ratio in unknown category and lowest A:E ratio in Taxable category
   * May have been data submission issue
   ▪ 2019 study – highest A:E in unknown category, but results and relationships now fit expectations

Question 8

Reliance and Limitations

Answer 8

• Relied on contributing companies for accuracy and completeness of data o Tested for reasonableness
  o Interpreted and adjusted as needed
• No formal audits of data
  o Relatively cursory review of experience
• Users should understand limitations of the data

Question 9

Next Steps

Answer 9

• Develop recommendations to make future iterations more efficient
• Make more pivot tables available
• Conduct in-depth studies of data and potential implications (especially SS data)